18th Dec 2025
AI-generated images are increasingly the starting point for physical production briefs. The visual is clear, the intent is obvious, and the question is always the same: can you build this? The answer is usually yes. Getting there requires a specific kind of translation.
A different kind of brief
For most of the history of physical production, briefs arrived as drawings, CAD files, sketches or physical references. They communicated intent through a language that designers and makers had developed together over decades. Even a rough sketch carries information about structure, scale and material that a trained eye can read.
AI-generated images communicate differently. They describe how something should look with remarkable precision while saying nothing about how it should be built. The surface finish, the proportions, the mood, the sense of weight and materiality, all of these can be conveyed with extraordinary clarity. The structural logic, the material behaviour, the construction sequence, the tolerances that will determine whether the piece holds together: none of these exist in the image.
This is not a criticism of AI as a design tool. It is simply a description of what the translation requires.
What the translation actually involves
When we receive an AI-generated visual as a brief, the first task is interpretation. We are reading the image the way a structural engineer reads an architect’s rendering: asking what would actually have to be true about this object for it to exist in three dimensions, behave as intended, and hold up to whatever use or scrutiny it will face.
Some elements translate directly. A form that looks like it could be CNC-machined from a solid block probably can be. A surface finish that reads as painted metal probably is. These decisions are straightforward.
Others require intervention. AI images frequently generate forms that assume materials behave in ways they do not: surfaces that appear to float without support, geometries that would require a wall thickness too thin to hold structural integrity, junctions that look clean but would trap stress in the real object. The job is to identify these elements, develop solutions that preserve the visual intent, and communicate the decisions clearly before production begins.
The Predator Chair
A client approached us with an AI-generated image of a chair. The design was aggressive and highly stylised, with a silhouette that read somewhere between industrial machinery and high-end furniture. The image communicated mood, proportion and surface quality with real clarity. It offered no information on structure, materials or how any of it would actually be supported.
Our task was to take that visual intent and build something that looked like the image and could also function as a chair. That required analysing the AI output carefully, understanding which elements were load-bearing versus decorative, developing a structural approach that the image did not contain, and selecting materials that would produce the right surface qualities at the right cost.
The resulting piece was faithful to the original visual while being engineered to actually work. The client had something they could sit in, photograph and present. The AI provided the spark; the fabrication required everything else.
Working with AI references
If you are commissioning physical production from an AI-generated image, the most useful thing you can do is be clear about which elements of the image are non-negotiable and which are open to interpretation. The proportions? The surface finish? The specific material? Knowing where you have flexibility and where you do not allows us to focus the development process on what matters most.
The other useful thing is to think about how the piece will actually be used. A photography prop has different requirements from an installation that will survive a month of public interaction. An AI image does not know the difference. We do, and it shapes every decision about materials and construction from the start.
AI is a powerful new input to physical production. It changes how ideas are communicated and how briefs arrive. What it does not change is the knowledge, craft and problem-solving required to turn a compelling image into a thing that exists.